To understand why once-weekly isoniazid/rifapentine therapy for tuberculosis was less effective than twice-weekly isoniazid/rifampin, we studied human immunodeficiency virus-seronegative patients with either failure (n = 4), relapse (n = 35), or cure (n = 94), recruited from a comparative treatment trial. In multivariate analyses that were adjusted for severity of disease, low plasma concentrations of isoniazid were associated with failure/relapse with once-weekly isoniazid/rifapentine (median isoniazid area under the concentration-time curve for 12 hours after the dose [AUC(0-12)] was 36 microg x hour/ml in failure/relapse versus 56 microg x hour/ml in control cases p = 0.005), but not with twice-weekly isoniazid/rifampin. Furthermore, two patients who relapsed with Mycobacterium tuberculosis monoresistant to rifamycin had very low concentrations of isoniazid. Finally, isoniazid acetylator status determined by N-acetyltransferase type 2 genotype was associated with outcome with once-weekly isoniazid/rifapentine (p = 0.03) but not twice-weekly isoniazid/rifampin. No rifamycin pharmacokinetic parameter was consistently and significantly associated with outcome (p > 0.10). Because low isoniazid concentrations were associated with failure/relapse, a drug with consistently greater area under the concentration-time curve than isoniazid may be needed to achieve highly active once-weekly therapy with rifapentine.